An OpenStudio Model (What is the Openstudio Model?) represents a single building energy model, either complete (simulatable with EnergyPlus) or partial. It is composed of a collection of ModelObjects (What is a ModelObject?), which are polymorphic, that is, ModelObject is the base class from which more specific object types are derived. Each leaf of the inheritance tree wraps a specific data object type that is ultimately defined by the OpenStudio.idd file. The following classes form the foundation of the OpenStudio Model:
The specific model object types can largely be broken down into the categories of simulation settings; output data; resources; site and location; geometry; building loads; advanced daylighting; heating, ventilation and air conditioning (HVAC) systems; economics; and standards. We will now list the most important classes in each of these categories. The lists are not meant to be exhaustive, but rather to give an intuitive sense of what is and is not currently covered by the OpenStudio Model. Throughout, some of the classes will be concrete (have a one-to-one mapping with an object in OpenStudio.idd), and some will be abstract (provide a higher level interface to multiple Idd objects).
This subsection of the model is not fully developed. Classes like ComponentCost_LineItem are available, but we expect to do a full refactor of this area of the model as time allows.
This subsection of the model is not fully developed. BuildingStandardsInformation, ConstructionBaseStandardsInformation, and ClimateZones are examples of this type of object.
The following links may also be of general interest:
The OpenStudio Model is a hierarchical, object-oriented model of a partial or complete building that exists to support annual energy and other simulations and related analyses. Complete building models may be simulated with EnergyPlus, which is developed by the Department of Energy and is our primary simulation target (http://apps1.eere.energy.gov/buildings/energyplus/). Daylighting simulation with Radiance is also supported, and OpenStudio can conduct some economic analyses natively. The OpenStudio Model aims to integrate input and output data into a seamless whole. EnergyPlus output data is accessed through the Model by connecting to an appropriate SqlFile (see the utilities documentation), which provides a basic interface to the EnergyPlus SQLite output. Once the basic connection to an output file is established, intuitive interfaces to specific pieces of output data are provided at the object level. For instance, see Facility, ThermalZone, and Output Data Access Methods.
The underlying input data for Model is adapted from the EnergyPlus IDD/IDF file formats, where an IDD (input data definition) file is a data schema, analogous to an XML Schema file, and an IDF (input data file) is an instantiation of that schema, analogous to an XML file. The OpenStudio IDD has been developed using a subset of objects in the EnergyPlus IDD as a starting point so that OpenStudio can directly leverage EnergyPlus's numerous and extensively documented data objects. Over time the OpenStudio IDD has grown to include more objects from the EnergyPlus IDD, with a focus on including the most important types for energy simulation first. Additionally, new object types have been added to the OpenStudio IDD to provide functionalities such as:
Models are serialized to files with the extension '.osm', which are instance files of the OpenStudio data schema. OSM files are clear text and look very similar to IDF.
The last thing the building energy modeling community needs is yet another data model. So why did we create the OpenStudio Model?
To harness object-oriented programming techniques to provide a powerful interface for building energy modeling at all fidelity levels, for various software development and user communities.
By formalizing relationships among data objects (Relationships); and by placing high-level, big-knob methods in appropriate data objects (Compound Methods); by making it possible to access related input and output data from the same location ( Output Data Access Methods); and by supporting multiple types of simulation engines, the OpenStudio Model provides a number of advantages over direct manipulation and analysis of EnergyPlus input and output formats.
Model serves as a container for ModelObjects and a SqlFile. Model inherits its functionality for storing and querying objects from openstudio::Workspace, which is essentially an in-memory, custom database for openstudio::WorkspaceObject. The hierarchy of object types will be described in full in Implementation Details. For now it is important to note that all of Workspace's methods (listed in the utilities documentation) are also available to Model. Ideally, the underlying Workspace methods should only be called in model .cpp files; Model and ModelObject provides methods for adding and removing objects with more built in intelligence. However, the basic getters (e.g., getObjectByNameAndReference) and query methods (e.g., numObjects, numObjectsOfType) may be of general interest.
Model is implemented using the pImpl (pointer to implementation) idiom. In this pattern, each public class (e.g. openstudio::model::Model) holds a pointer to an implementation class (e.g. openstudio::model::detail::Model_Impl) that actually holds the object data and implements its methods. The public classes forward almost all user requests to the implementation class. This idiom was adopted because it results in data sharing between copies of public class instances, without the need for exposed pointers; and because it provides consistency of interface across native C++ and the SWIG bindings. (At this time, the OpenStudio SDK is available in Ruby and C#, in addition to C++, by way of SWIG.)
Example of data sharing:
At its most basic, ModelObject is the OpenStudio class that represents an IDF object. Just like a text IDF object, it consists of a number of ordered string fields, some of which may be in extensible groups, and all of which conform to the corresponding IDD object. Please see the utilities IDD and IDF documentation (specifically openstudio::IddObject, openstudio::ExtensibleIndex, and openstudio::IdfObject) for more information. ModelObjects are always owned by/live in a containing Model, which makes it possible for ModelObjects to point to each other in a meaningful way, even when multiple Models are being operated on side-by-side.
Although, strictly speaking, ModelObject provides access to its underlying string fields, a user of the OpenStudio SDK should not need to resort to that low-level interface in most circumstances. ModelObject is the base class for a large inheritance hierarchy, the terminating nodes of which each wrap a specific object in the OpenStudio IDD. Each data field that should be publicly accessible is wrapped into a getter and a setter method (see Field Getters and Setters), several types of Compound Methods and Output Data Access Methods are provided, as are object getters and setters appropriate to the various Relationships in the Model. Methods for creating and deleting data are also provided and available through a uniform interface, see Key Behaviors.
At this time, ModelObjects are uniquely identifiable by their IddObjectType and name. ModelObjects also have a unique handle that is available and used in memory, but is currently not serialized to file. A number of objects are unique in a given model. Unique objects may or may not have a name, as a name is not strictly necessary for identification. For example:
ModelObjects of different types can share a name as long as they are not be in the same IDD reference list; ModelObjects of the same IddObjectType cannot.
Every data field that the user should be able to directly manipulate (possibly with side effects, as in Lights::setLightingLevel), is exposed through getter and setter methods tailored to the specifics of that field as specified by the IDD. For instance, the IDD slash code '\type real' results in getters and setters on double or boost::optional<double>. The name of the field is generated off of the IDD field name, keeping the names as close to the same as possible, but removing non-word characters, and transforming to lower camel case. Fields that point to other objects (\type object-list) are not exposed in quite the same way, see Relationships.
By maintaining the field names defined in the IDD, and by listing the name of the IDD object wrapped by a given ModelObject in each concrete object's documentation, we maintain a link with the EnergyPlus Input-Output reference for those ModelObjects that directly or essentially wrap a particular EnergyPlus IDD object. We therefore do not provide detailed documentation for these methods, unless the field was introduced by the OpenStudio IDD. However, side effects, like multiple fields being touched by one setter, are documented.
Data fields in extensible groups also have getters and setters, but may be exposed in different ways, for instance, in a getter that takes the extensible group's index as an argument (e.g., GasMixture::getGasType, GasMixture::setGasType), as part of a class that derives from ModelExtensibleGroup (e.g. ClimateZone), or through the introduction of a specialized data structure (e.g. PlanarSurface::vertices, PlanarSurface::setVertices). Classes that derive from ModelExtensibleGroup are generally preferred, as they let the extensible groups be dealt with on their own terms as individual objects. An exception to this is the case where there are just a few simple pieces of data in the group, then a small struct (like the one used by PlanarSurface::vertices) may be preferable. A ModelObject with a companion ModelExtensibleGroup should provide a wrapper method for numExtensibleGroups with the name of the ModelExtesibleGroup derived class replacing "ExtensibleGroups", and similar wrappers for getExtensibleGroup, extensibleGroups, clearExtensibleGroups and pushExtensibleGroup (perhaps replacing "push" with "add"), at a minimum.
In general, we make the following translations from IDD slash codes to C++ syntax.
IDD fields that are '\required' are provided in the object constructors, and are always returned by value (double, int, or std::string, as opposed to boost::optional<double>, boost::optional<int>, or boost::optional<std::string>, respectively), unless the field can accept different data types (as is the case with '\autocalculatable' and '\autosizable' fields). IDD fields with '\default' values are also returned by value if the field can only accept one data type (e.g. it is not '\autocalculatable' or '\autosizable'). Note that some fields in the EnergyPlus IDD are marked as '\required' and '\default', this is logically inconsistent so these fields are treated as if they were marked only as '\default' and not as '\required'. It is preferable to mark fields as '\default' in the OpenStudio IDD if there is a suitable default value instead of providing default arguments to the ModelObject constructor.
If, the field is not required and does not have a default or the field can contain multiple data types (e.g. such as '\autocalculatable' or '\autosizable' fields), the return value is wrapped in a boost::optional. If such an optional return value evaluates to false, then there is no underlying data, and the user must not dereference the optional container. If the return value evaluates to true, the actual returned data may be accessed by dereferencing with '*' or '->', for example:
If the user goes around the Model getter and setter methods by hand editing files or using the WorkspaceObject interface directly, it is possible to break the return-by-value getters, in which case a fatal error will be logged, and an openstudio::Exception will be thrown. To make it clear to a user when a default value is set (the underlying field is empty), fields with default values are provided with a query bool [fieldName]IsSetToDefault() const.
If it is possible for a setter to fail, the setter will return a boolean to indicate success (true) or failure (false). Otherwise, the return type will be void, and the user can trust that the specified action will execute as expected.
Numeric fields are based on the C++ data types double (if '\type real') or int (if '\type integer'). The getters return a value if the field is required or has a numeric default, and is not '\autocalculatable' or '\autosizable'; and return an optional otherwise. The setters accept a double or int depending on the appropriate type. If the IDD specifies a minimum or maximum value then setter may fail and thus returns a bool; otherwise the set operation cannot fail and void is returned. If the field has a default, the user can check if the field is defaulted using [fieldName]IsSetToDefault, the user can set the field to default value using set[FieldName]ToDefault. If the field has can be autocalculated, the user can check if the field is autocalculated using [fieldName]IsSetToAutocalculate, the user can set the field to autocalculate using set[FieldName]ToAutocalculate.
Fields of '\type choice' are IDD defined enumerations. Originally, enum classes were used for choice fields. However, those classes greatly increased the size of the Model library. Therefore, the following rules were developed.
If the IddKeys associated with a choice field are (case-insensitive) 'Yes'/'No', 'On'/'Off', or 'True'/'False'; then the OpenStudo getters and setters use bool values. Two additional methods 'std::string [fieldName]String() const' and 'bool set[FieldName]String(const std::string&)' are provided if the user needs to work with the string values. The general rules concerning getter return values, and setter return values apply to string versions of the field getters and setters. The general rules concerning default fields apply to the bool version of the getter and setter. However the bool setters cannot fail, and so always return void. Finally, if there is no default value and the field is not required the bool getter will return a boost::optional<bool> the user must be careful not to confuse whether this optional is initialized with the actual value of the field.
Other choice fields operate solely off of std::string. The allowable values are case insensitive, defined by the IDD and exposed through static methods like Gas::validGasTypes(). The setters can always fail; whether a boost::optional<std::string> is returned by the getter or accepted by the setter conforms to the general rules.
The OpenStudio IDD introduces the '\type url' field to help with locating files other than the OSM on which a simulation depends. (See WeatherFile for instance.) RunManager provides functionality to normalize url fields at job creation time. RunManager looks for the file across all URLSearchPaths (see utilities documentation) passed in to the job creation. In this process, if the referenced file can be found, RunManager ensures that the file is copied to the local computer running the job and that the url field is set to a valid absolute path to that file for all child jobs (the url field is unchanged in the original input Model). If the field immediately following a url field is named 'Checksum' and is of type alpha, that field contains a checksum (computed by openstudio::checksum()) that RunManager uses to verify that the file found is the desired file. External files may be referenced using url fields in several ways:
Url fields are implemented with two overloaded set methods which each return void; one taking a string and one taking an openstudio::path. The get method returns a string. Normal rules for required and default fields apply.
All other types of fields ('\type alpha' or '\type external-list') are treated as strings and conform to the general rules.
An extended example of data field getter and setter usage:
OpenStudio ModelObjects provide "big knob" compound methods to enable high-level modeling and various types of analysis (perturbation, optimization, standards, etc.) For example, Space::lightingPowerPerFloorArea calculates lighting power in the space by adding up lighting power from all Lights and Luminaire objects directly assigned to the space or it's space type, then dividing by the floor area of these space. Furthermore, that method, and the Space compound "getter" that corresponds to it are wrapped up together in an Attribute named "lightingPowerDensity" available through the ModelObject::getAttribute interface.
PlanarSurface::uFactor and PlanarSurface::setUFactor (and the associated Attribute) is another example that brings a lot of pieces together to expose a high-level piece of data. Depending on the underlying construction type, these methods may operate directly on a SimpleGlazing object; may use static film resistance data in conjunction with one or more Material objects, and maybe a ConstructionBaseStandardsInformation object (the latter is sometimes needed for the u-factor setter); or may query the output data in the SqlFile. If u-factor is available through input and output data, the output data is returned, but in the meantime the two values are compared and a message is logged if their difference exceeds some tolerance. Related methods are available through ConstructionBase.
Several objects include a number of methods (some static) provided to give the user access to basic physical information used in EnergyPlus or in building energy standards. The film resistances alluded to above are available through PlanarSurface::filmResistance and PlanarSurface::stillAirFilmResistance. Physical properties of gases commonly found in multi-paned windows are available from FenestrationMaterial.
Overall, the compound methods are difficult to characterize as a whole. We encourage users to peruse the documentation for each individual object in which they have an interest. Compound methods are generally listed outside of the "Getters" and "Setters" sections, even though some methods' functionality will be similar to a get or a set. In general, these methods should be well documented since they are native to OpenStudio.
The model.setSqlFile(const SqlFile& sqlFile) method is used to attach a Model to related EnergyPlus simulation output. Because the OpenStudio Model must be translated to EnergyPlus IDF for simulation, the results of the EnergyPlus simulation will not map exactly to the Model. However, the Model will perform some checks to verify that the given SqlFile does relate to the given Model, the setSqlFile operation will fail if these checks are not satisfied. If model.sqlFile() evaluates to true, then a given model is hooked up to its EnergyPlus simulation output data in the form of a SQLite file, and a number of ways to retrieve output data become active.
The connection to the SqlFile can be reset using model.resetSqlFile(). Closing the SqlFile connection is necessary in order to resimulate the Model using EnergyPlus.
The Parent-Child relationship is used to represent containment relationships in the building. For example, Building parents Space, which parents a number of objects including Lights, People, and ElectricEquipment. If a ParentObject is removed, its children (and on down the line recursively to grandchildren, etc.) should also be removed; and if a ParentObject is cloned (copied), its children (recursively) should also be cloned.
In addition to helping define behaviors, the Parent-Child relationship is useful for model visualization and navigation. Navigating using ModelObject::parent() and ParentObject::children() in a Ruby script, for instance, can give the user intuition about how the various aspects of a Model conceptually fit together in OpenStudio and EnergyPlus.
A ResourceObject is a ModelObject that can be used by multiple other data objects. For instance, unlike a Lights object which can only be associated with one Space or SpaceType, a Schedule can be used by any number of objects, so long as the data type of the Schedule (fractional, on/off, temperature, etc.) is appropriate for the context. Every ModelObject can return the ResourceObjects that is uses with the method ModelObject::resources(). ResourceObjects can report their use counts, see ResourceObject::directUseCount and ResourceObject::nonResourceObjectUseCount. Unused resources can be removed in one fell swoop with Model::purgeUnusedResourceObjects.
Unlike children, resources are meant to be shared, so they are not removed when a user of the resource is removed, and they are not automatically cloned when one of their users is cloned (see Key Behaviors).
Parents, children and resources are available through generic and typed interfaces. For instance, Building::children() will return all of the Spaces in a std::vector<ModelObject>, and Building also has a method spaces() that returns the same data as a std::vector<Space>. The Schedule associated with a Lights object can be found through the std::vector<ResourceObject> returned by Schedule::resources(), and also through the method Lights::schedule(). Lights also has the method setSchedule(const Schedule&) .
Connections join multiple ModelObjects together via designated points called ports. Connections between ModelObjects are directional and they are created by identifying a source ModelObject and port as well as a target ModelObject object and port. Once a connection is made, it is easy to access the ModelObject connected to a particular port using ModelObject::connectedObject. New Connections are made using the openstudio::model::Model::connect function.
ModelObject connections are similar to EnergyPlus nodes, however a connection between ports is more general than a node connection in Energyplus. In the OpenStudio model, connections serve the same role as nodes in EnergyPlus, but OpenStudio connections can be applied in contexts other than the EnergyPlus system nodes. OpenStudio's connections and ports can be used any time one ModelObject can be arbitrarily connected to another ModelObject. For example, EnergyPlus EMS sensors and actuators could be thought of as source and target ports, and a connection could be made between the two.
As described above, objects in the OpenStudio Model can be organized into a tree (actually a forest) by linking ModelObject types based on their Parent-Child relationships. The following is a complete listing of the concrete ModelObjects displayed in Parent-Child hierarchical form. The immediate base class of each object is listed in parentheses. Some object types are listed more than once because they can be the child of multiple concrete types (in which case the Parent-Child relationship is often defined by an abstract base class). If the class name is somewhat different from the IddObjectType valueName, the name of the IDD object is also listed for reference.
Special Cases that can be Parented by many ModelObjects
There are four constructors available for Model. The default constructor creates a new Model with a single Version object in it. Version objects are not returned by Model::objects or counted by Model::numObjects.
The preferred method for creating a Model from an OSM file is
UNLESS that OSM file might be from an old version of OpenStudio, in which case the version translator should be used:
The preferred method for creating a Model from IDF file is
In addition, constructors are provided from IdfFile and Workspace (see the utilities documentation). Those constructors will throw if IdfFile or Workspace uses the wrong IDD file.
Because Model is required to support partial (unsimulatable) building models for data sharing, and to provide a flexible interface for editing, it is ultimately up to the user to ensure that a Model is ready for simulation. Model::validityReport and similar methods (see the utilities documentation, especially Workspace::validityReport, WorkspaceObject::validityReport, and ValidityEnums.hpp) are provided to help users bring their models up to the standard set by the IDD. (And there are a few places in Model where validity checking goes beyond the data schema to rules internal to EnergyPlus, see for instance LayeredConstruction::layersAreValid, which is used by the virtual implementation of validityReport for LayeredConstructions .) Beyond that, users will sometimes need to go through an iterative process of modeling, trying to run EnergyPlus (or another simulation engine), addressing simulation errors, etc.
Every concrete ModelObject has one or more public constructors, unless it is marked unique by the IDD. In this case, there is no public constructor in order to prevent the accidental construction of multiple instance of that type in the same model. Instead, these objects are made available through Model::getUniqueModelObject and Model::getOptionalUniqueModelObject:
For non-unique objects, all public constructors will ensure that required fields, children, and resources are present as a post-condition. Depending on the constructor, the user may be required to explicitly provide those objects, or they may be default constructed. For example the LightsDefinition object is a required argument to the Lights object constructor. Parents are often not required by the public constructors; as appropriate, convenience constructors that take the parent as an argument may be provided. (Orphan construction is allowed in order to better support data sharing. For instance, if someone wants to create a lighting component for the Building Component Library, requiring Lights to be constructed with a Space or SpaceType would be inconvenient.)
ModelObject::remove is used to remove a ModelObject from a Model. When a model object is removed, all of its children are also removed. Objects related to the removed object through a resource or connection relationship are not deleted. Resources persist so they can be used by other objects. Connections are broken by object removal, but the adjacent object is left behind, unless it happens to also be a child of the removed object. However, when a resource or child object is removed, its user objects or parents may also be removed, if the relationship is a required one for the user/parent. For instance, SpaceLoadDefinition::remove also removes all of its SpaceLoadDefinition::instances, since a SpaceLoadInstance cannot exist without its definition.
For all deleted objects, when all the instances of that object go out of scope and stop referencing the shared implementation pointer, the memory will be automatically freed for you.
ModelObject::clone is used to duplicate a ModelObject. The clone method may be used to copy an object into another model, or to duplicate an object in the same model. The act of cloning an object will change the duplicate object's name if an object of the same type and name already exists in the target model. Therefore, cloning into the same model will always result in name change for the duplicate so as to avoid a name collision with the original object. Cloning a ModelObject also clones all of its children (recursively). Resources are cloned as necessary (when an equivalent object cannot be located, see IdfObject::dataFieldsEqual and IdfObject::managedObjectListFieldsNonConflicting in the utilities documentation) when the original and target models are different.
The following code would also pass its OS_ASSERTs. Because we cloned into the same model, the schedule (a ResourceObject) was NOT cloned.
Similar to Model inheriting from openstudio::Workspace, ModelObject inherits from openstudio::WorkspaceObject, which in turn inherits from openstudio::IdfObject. Briefly, IdfObject corresponds to a single data object in an IDF file. It has an IddObject, which defines its type, and it otherwise consists of a vector (ordered list) of string fields, accessible by index and with an optional user comment. (If you never specify an IDF field comment, default comments are generated for you from the IddField::name()'s.) WorkspaceObject provides the added functionality of turning the '\type object-list' fields that in clear text point to other data objects by name reference into bi-directional, name-independent pointers to particular objects. Please see the utilities/idf documentation for more information.
In this way, IdfObject, WorkspaceObject, and Workspace form the underlying data model used by OpenStudio to represent a single building (assuming the OpenStudio, rather than the EnergyPlus or other, IDD is referenced). Model, ModelObject and everything else in the openstudio::model namespace exists to turn that data model into a full-featured object model.
Because Model and ModelObject publicly inherit from Workspace and WorkspaceObject, methods of Workspace and WorkspaceObject are directly available for use through Model and ModelObject. Use of these methods is discouraged where it is not necessary as these data model-level modifications bypass the best-practice logic built into classes derived from ModelObject.
The pointer to the implementation is managed by a reference counting pointer, so the implementation object is destroyed when the last wrapper class goes out of scope. Because the implementation is shared between copies of the public class, signal and slot connections will remain active after the public object is destroyed provided that another object still references the same implementation. Note that at the very least this other object is typically the Model of which the ModelObject is a member, as Model (actually Workspace) maintains a map of all of its objects' implementation reference counting pointers.
The combined use of template methods such as Model::getModelObjects<T>() with the pImpl idiom has led to the unfortunate case where users of a class may need to also include the corresponding implementation class, for example:
Another important detail about the OpenStudio Model is that ModelObjects are only valid when they are part of a Model. The benefit of this requirement is that any ModelObject can access the entire Model to search for related ModelObjects. For the most part, the Model interface enforces this requirement. However, there are some cases where it is possible to break this invariant, for example:
A Component is a partial Model that can be created from a ModelObject and then shared between models or with other modelers. When ModelObject::createComponent is called, ModelObject::clone is used to copy that object (the primary object) and related objects (often the children and resources of the object, evaluated recursively) into a new, blank Component. A Component is a Model that satisfies a few additional constraints.
Components always contain one ComponentData object, which provides a place to list all the objects in the Component, as well as universal identifiers and timestamps used to keep track of the Component as it travels to and from local and remote Building Component Libraries (BCLs). The contents list is always headed by the primary object, which induces a parent-child tree plus resource and connection relationships on the rest of objects. For instance, a Construction Component would have a Construction object as its primary object, and would likely also contain a number of Material objects (resources of the Construction), and a ComponentCost_LineItem object (child of the Construction).
A component creation example follows:
The Component class disallows some of the non-const operations of Model. In particular, no objects can be added to or inserted into a Component. Objects can be removed, except for the primary ModelObject and the ComponentData object. If either of those objects are removed, the Component is invalidated and throws an exception. Once created, Components can be saved for future use, or directly used in another Model. To save a Component for future use, it should be added to a local BCL, and/or uploaded to the online Building Component Library. (These features are currently under development. Please see the documentation for utilities/bcl and http://bcl.nrel.gov/.)
To use a Component (either directly, or deserialized from a building component library) in a Model, use the method Model::insertComponent(const Component& ). This method first looks to see if a copy of the Component is already present in the Model; if not, it clones all of the objects into the Model. In either case, if successful, the method returns the resulting ComponentData object. After an insertComponent operation, Model watches the contents of the inserted Component, changing the ComponentData::versionUUID if data is modified, and deleting the ComponentData object (and discontinuing the watching operation) if it is invalidated (by removal of the primaryObject, for instance). In this way, Model can be queried to find out what Components it uses, and one can ascertain whether any data has changed as compared to the original version. Therefore, this is a mechanism that can be used to trace coherent pieces of input data through a series of related models, between similar research project models, or to ensure that the appropriate components are used for building energy standards compliance.
An example of using the component created above: